Korg Forums Forum Index Korg Forums
A forum for Korg product users and musicians around the world.
Moderated Independently.
Owned by Irish Acts Recording Studio & hosted by KORG USA
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Creating a new multi-layer Steinway Piano for Pa2x/Pa800
Goto page 1, 2, 3, 4, 5, 6, 7, 8  Next
 
Post new topic   Reply to topic    Korg Forums Forum Index -> Korg Pa2XPro - Pa800 - Pa500
View previous topic :: View next topic  
Author Message
Rob Sherratt
Platinum Member


Joined: 16 Apr 2007
Posts: 4536

PostPosted: Fri Feb 13, 2009 9:41 am    Post subject: Creating a new multi-layer Steinway Piano for Pa2x/Pa800 Reply with quote

The Steinway Piano Project

Something that I found helpful when working on a new multi-layer 88 note Steinway piano sound for the Pa2x/ Pa800 was to have a plan that showed me exactly which sample was assigned to which key, the note frequencies, the octave numbers etc. This plan might be helpful to other forum members as well. Here is the picture so you know what I'm talking about. The bright yellow notes are the ones that have exact samples, then each of those samples is stretched by a semitone each way, with the exception of the B0 samples which have to stretch down by 2 semitones. Three new multisamples, Steinwaypp, Steinwaymf and Steinwayff are being constructed as you can see from the plan.



If you would like a copy of the .xls spreadsheet please send me an EMAIL (not a PM) by clicking on the EMAIL button below.

The Steinway piano project is something I wanted to do for a long time, and Reuben (John) pointed me in the direction of the University of Iowa Electronic Recording Studios, who have published 264 high quality stereo .aiff samples of their Steinway Grand Piano, each one up to 10 Mbytes in size. I wrote to the University and the department head Mr Lawrence Fritts has given consent to use these samples to make a Steinway Piano free give away to all Korg Forum members. Here is the copyright release statement he sent me in email on 12th Feb 2008:

Lawrence Fritts wrote:
The Steinway Piano samples were recorded in November 2001 by the University of Iowa's Electronic Music Studio department. http://theremin.music.uiowa.edu/MIS.piano.html The samples are made available for non-commercial use on the Korg Pa series keyboards, by permission of the department head, Mr Lawrence Fritts.

Would anyone want a sound using 2.64 Gbytes sample RAM? I expect not, so the first thing I have to do is to reduce the total sample size down to 63 Mbytes by using various tricks of the trade that I am learning as I go. It will fit easily on a Pa2x but Pa800 owners will have a hard choice to make I'm afraid, the Steinway samples will be all you have space for.

The reason for the large sample set is that it will be a three-layer sound (ie using different recordings for pp, mf and ff which will be velocity layered and cross faded using a technique Nikola taught me). It will use authentic samples for attack and decay, although I will be looping the tailpieces and then creating a software based sustain and release envelope so that I can get a longer sustain and delay effect than the original Steinway when needed. There will be several sound variants with different sustain, decay and EFX parameters.

Please don't hassle me for a completion date, it is a lot of work and I'm doing it as a hobby and I want to take my time to get it right. It is also the first time I've constructed a complete high quality piano sound from scratch, so I've been relying very much on Sharp's help behind the scenes on the best way to do it.

If anyone can recommend any ways to automate the sample editing/ preparation stages (i.e. topping and tailing at a 10 sec cutoff point, and reducing recording hiss, working with .aiff 16 bit 44.1 kSamples/sec on the PC) please send me an EMAIL. At the moment I am editing each sample individually and it takes a LOT of time to get it right!!

There will be a complete Wiki tutorial on creating the Steinway piano in due course. If people want me to, then I can publish follow-up messages and pictures on this thread as each stage of the project and the corresponding write-up is completed.

Best regards,
Rob
Back to top
View user's profile Send private message Send e-mail
mrkorg
Platinum Member


Joined: 03 Nov 2007
Posts: 548
Location: Australia

PostPosted: Fri Feb 13, 2009 2:00 pm    Post subject: Reply with quote

Fantastic, very good. I thought Korg can only use 1 layer (at least this is what Nikola told me).
_________________
KORG PA2XPRO
Back to top
View user's profile Send private message
Sharp
Site Admin


Joined: 02 Jan 2002
Posts: 17929
Location: Ireland

PostPosted: Fri Feb 13, 2009 2:17 pm    Post subject: Reply with quote

mrkorg wrote:
Fantastic, very good. I thought Korg can only use 1 layer (at least this is what Nikola told me).


Nope, you can use up to 16 layers.

Regards.
Sharp.
_________________
KORG Store - Irish Acts Irish Acts Online Store
Back to top
View user's profile Send private message Send e-mail Visit poster's website
BasariStudios
Approved Merchant
Approved Merchant


Joined: 29 May 2005
Posts: 6507
Location: NYC, USA

PostPosted: Fri Feb 13, 2009 3:02 pm    Post subject: Reply with quote

Rob, i want a 2.6GB Piano...i am already using one of 1.6GB Piano.
...in Kontakt...
_________________
http://www.basaristudios.com
Cubase 8.5 Pro. Windows 7 X64. ASUS SaberTooth X99. Intel I7 5820K. ASUS GTX 960 Strix OC 2GB. 4x8 GB G.SKILL.
2 850 PRO 256GB SSDs. 1 850 EVO 1TB SSD. Acustica: Nebula Server 3 Ultimate, Murano, Magenta 3, Navy, Titanium.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Rob Sherratt
Platinum Member


Joined: 16 Apr 2007
Posts: 4536

PostPosted: Fri Feb 13, 2009 3:31 pm    Post subject: Reply with quote

Funny enough, Nikola and I had been discussing it on another thread. You can create velocity triggered multilayers which cross fade between different sample sets. I think it's done as follows:

a) First pair of osscillators will handle the stereo SteinwayPP multisample
b) Second pair will handle the stereo SteinwayMF multisample
c) Third pair will handle the stereo SteinwayFF multisample.

Then I will set up the velocity trigger parameters something like this

a) range 2-40, max gain output = 80
b) range 41-80, max gain output = 110
c) range 81-127, max gain output = 127

Now comes the clever bit that Nikola told me about, we use another two pairs of oscillators to do cross fading as follows:

d) same multisamples as (a) range 41-80, max gain = minus 80
e) same multisamples as (b) range 81-127, max gain = minus 110

If I understood Nikola correctly that means that the PP samples will play at a velocity of 41 with an output gain of 80, and will taper down to an output gain of 0 with an input velocity of 80. Hence they will be layered with the MF samples. And a similar case for the MF and FF samples. At any one time there will be four OSCs running so this will be a little heavy on polyphony.

And the remaining OSCs will be used for sustain and pedal RX noises like the Grand Piano RX which I will use as a template.

If anyone has done this before, please let me know if my approach is correct? It may save me a lot of time!!

Best regards,
Rob
Back to top
View user's profile Send private message Send e-mail
Rob Sherratt
Platinum Member


Joined: 16 Apr 2007
Posts: 4536

PostPosted: Fri Feb 13, 2009 3:44 pm    Post subject: Reply with quote

BasariStudios wrote:
Rob, i want a 2.6GB Piano...i am already using one of 1.6GB Piano.
...in Kontakt...


The Steinway samples from Iowa University are free for non commercial use, so you could ask them ... it would be easier for you or me to create a 2.6 Gb piano than a 63 Mb piano, just a lttle more expensive in sample RAM!! I guess the Oasys could handle it ...

Best regards,
Rob
Back to top
View user's profile Send private message Send e-mail
a.schemkes
Senior Member


Joined: 06 Jul 2008
Posts: 355
Location: Netherlands [PA2xPro + M3-61]

PostPosted: Fri Feb 13, 2009 8:32 pm    Post subject: Reply with quote

Wow are you crazy Rob? Is there no way that we can use the new 128Mb sample that's available for the M3????
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Rob Sherratt
Platinum Member


Joined: 16 Apr 2007
Posts: 4536

PostPosted: Fri Feb 13, 2009 8:42 pm    Post subject: Reply with quote

a.schemkes wrote:
Wow are you crazy Rob? Is there no way that we can use the new 128Mb sample that's available for the M3????


Hi Anton,

Which 128 Mb sample is that? There is the EXBUSB03 new piano set which contains many piano samples, each about 10 Mb. Then there is the EXBUSB04 oldpiano which is 20 Mb. None of these sound "right" to me in the way a multi layered Steinway should sound. I don't think separate recordings were taken at pp, mf and ff, I just think there are separate ADSR envelopes applied to the same samples depending whether you play pp, mf and ff. That's what it sounds like to my ears, and that's how Korg implemented the Grand Piano EX on the Pa2x.

Also, I don't have the technical knowledge to transfer samples from the M3 to the Pa2x. Maybe someone else does? I think there is an incompatibility and file protection issue.

Anyway, I'm about halfway through editing the "one third" of the University of Iowa samples now. I switched to using a free editor called WavePad which is extremely good and is easier to use than the editor on the Pa2x.

Maybe I am crazy. I keep asking my other self that question Laughing

Regards,
Rob
Back to top
View user's profile Send private message Send e-mail
Sharp
Site Admin


Joined: 02 Jan 2002
Posts: 17929
Location: Ireland

PostPosted: Fri Feb 13, 2009 9:07 pm    Post subject: Reply with quote

Rob Sherratt wrote:
Funny enough, Nikola and I had been discussing it on another thread. You can create velocity triggered multilayers which cross fade between different sample sets. I think it's done as follows:

a) First pair of osscillators will handle the stereo SteinwayPP multisample
b) Second pair will handle the stereo SteinwayMF multisample
c) Third pair will handle the stereo SteinwayFF multisample.

Then I will set up the velocity trigger parameters something like this

a) range 2-40, max gain output = 80
b) range 41-80, max gain output = 110
c) range 81-127, max gain output = 127

Now comes the clever bit that Nikola told me about, we use another two pairs of oscillators to do cross fading as follows:

d) same multisamples as (a) range 41-80, max gain = minus 80
e) same multisamples as (b) range 81-127, max gain = minus 110

If I understood Nikola correctly that means that the PP samples will play at a velocity of 41 with an output gain of 80, and will taper down to an output gain of 0 with an input velocity of 80. Hence they will be layered with the MF samples. And a similar case for the MF and FF samples. At any one time there will be four OSCs running so this will be a little heavy on polyphony.

And the remaining OSCs will be used for sustain and pedal RX noises like the Grand Piano RX which I will use as a template.

If anyone has done this before, please let me know if my approach is correct? It may save me a lot of time!!

Best regards,
Rob


Hi Rob.
You might like to try this with existing ROM samples on the Pa2X before you spend lots of time on the layering. Proper programming of the filters to hide the exchange between sample layers is by far more transparent than blending layers in over each other without a proper cross fade system.

Your very likely going to hear chorus / phase problems that you might not like.

Regards.
Sharp.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
BasariStudios
Approved Merchant
Approved Merchant


Joined: 29 May 2005
Posts: 6507
Location: NYC, USA

PostPosted: Fri Feb 13, 2009 9:44 pm    Post subject: Reply with quote

I agree with James too and especially cuting that size down to only 64MB
i dont know...its a little bit scary.
_________________
http://www.basaristudios.com
Cubase 8.5 Pro. Windows 7 X64. ASUS SaberTooth X99. Intel I7 5820K. ASUS GTX 960 Strix OC 2GB. 4x8 GB G.SKILL.
2 850 PRO 256GB SSDs. 1 850 EVO 1TB SSD. Acustica: Nebula Server 3 Ultimate, Murano, Magenta 3, Navy, Titanium.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Lee
Platinum Member


Joined: 29 Dec 2004
Posts: 1198
Location: IN, USA(PA2XPRO)

PostPosted: Fri Feb 13, 2009 10:19 pm    Post subject: Reply with quote

Very interesting...I just this week started work on a 64MB Piano. I am using Kurzweil samples I have. They are triple strike (3 samples per note). The Kurzweil Triple strike piano is superb. What I don't know yet is how good it will be on the PA2 using these sampe conveted.

I may make a mess of it..but I'm determined to give it a best go!
If the results are good at all I can post a demo...but since these samples are sold (licenced) to me I can not share the sound with anyone else. Sad

BUT, I can share what I learn.

Rob, Sharp...thanks for all the heads up (hints) ...very timely!

OH, question for Sharp...you mentioned usng the ROM samples...How would we get a better (different) sound than Rob has already done with his Piano RX MOD using ROM sound?

Lee
Back to top
View user's profile Send private message Send e-mail
Rob Sherratt
Platinum Member


Joined: 16 Apr 2007
Posts: 4536

PostPosted: Fri Feb 13, 2009 10:38 pm    Post subject: Reply with quote

Hi Nedim and Sharp,

Thank you for your advice. I really appreciate the time you are taking to help me. What you say about unwanted chous/ flanger affect if I try to mix samples in a cross fade makes sense because I will not be able to get the sounds in phase alignment. So, I will abandon the cross fade OSC steps (d) and (e). It will increase polyphony anyway.

Question 1: Can you elaborate on using filters to blur the velocity triggered change from pp multisamples to mf, and from mf multisamples to ff? I know how to use filters for high/ low and band pass and I'm aware of the uses of filter resonance affects. But I cannot see how they will help in this specific case. Can you help me understand?

Question 2: I will be looping the tail portion of the samples and then using the Sustain-Release part of the ADSR envelope to shape the sustain and release portions. I believe that Attack should be a vertical line and Decay will be a horizontal line because I want the natural Attack and Decay of the samples to remain unaltered. Do you agree?

Question 3: There is a natural chorusing amplitude modulation in the samples, especially the bass notes, and I want the complex natural chorus in the samples to repeat when the notes are sustained. Here is a picture of the tail portion of one of the B0 samples, the highlighted region repeats naturally and it is that which I propose to loop when I load it into the Pa2x. Do you agree?



Question 4: I am working at the PC right now, normalising the ff samples all to 100% and the mf samples all to 75% and the pp samples all to 50%, and I can make other adjustments after they've been loaded onto the Pa2x. I was concerened not to raise the noise floor disproportionately by attempting to normalise the pp samples to 100% for example. Do these relative levels for the three sample sets seem reasonable to you?

Best regards,
Rob
Back to top
View user's profile Send private message Send e-mail
Sharp
Site Admin


Joined: 02 Jan 2002
Posts: 17929
Location: Ireland

PostPosted: Fri Feb 13, 2009 10:43 pm    Post subject: Reply with quote

Quote:
OH, question for Sharp...you mentioned usng the ROM samples...How would we get a better (different) sound than Rob has already done with his Piano RX MOD using ROM sound?


Using the layers in the configuration that Rob mentioned will cause chours and phase problems. I'm simply mentioning this in a round about way just to make Rob aware of what to expect, and to do some tests with the ROM sounds before spending any time make up 2 additional RAM based multisamples for something that might not work as he expects, or to a standard he will consider acceptable.

Regards.
Sharp.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
Sharp
Site Admin


Joined: 02 Jan 2002
Posts: 17929
Location: Ireland

PostPosted: Fri Feb 13, 2009 11:09 pm    Post subject: Reply with quote

Quote:
Question 1: Can you elaborate on using filters to blur the velocity triggered change from pp multisamples to mf, and from mf multisamples to ff? I know how to use filters for high/ low and band pass and I'm aware of the uses of filter resonance affects. But I cannot see how they will help in this specific case. Can you help me understand?


Great, your already thinking along the right lines already.

Considering that the frequency value can be modulated by the velocity, then it's only a matter of setting the filters up so that each Layer of samples is not fully open until you reach the top volume.

For example, if you only ever assigned 1 double forte multisample, then the sample should only be heard unfiltered when the key is pressed at volume 127.

If you set all this up correctly, the filters allow for an almost transparent switch between all 3 layers.

Quote:
Question 2: I will be looping the tail portion of the samples and then using the Sustain-Release part of the ADSR envelope to shape the sustain and release portions. I believe that Attack should be a vertical line and Decay will be a horizontal line because I want the natural Attack and Decay of the samples to remain unaltered. Do you agree?


Yeah, sounds good to me. The only thing that would worry me is the maths on the final file sizes. I don't believe you will fit all this into 64MB of memory if your waiting for the decay to level out so you can apply your loop.

That's why I mentioned using a Level in our PMs to each other. There's nothing wrong with your approach, it's just going to be hard fitting all this into 64MB.

Quote:
Question 3: There is a natural chorusing amplitude modulation in the samples, especially the bass notes, and I want the complex natural chorus in the samples to repeat when the notes are sustained. Here is a picture of the tail portion of one of the B0 samples, the highlighted region repeats naturally and it is that which I propose to loop when I load it into the Pa2x. Do you agree?


Yes the strings vibrate and pass through resonance cycle. For piano sounds that you have to loop this in order to sit back and let the Decay of the Synth engine do the rest. Otherwise your files will be masive altogether.

Quote:
Question 4: I am working at the PC right now, normalising the ff samples all to 100% and the mf samples all to 75% and the pp samples all to 50%, and I can make other adjustments after they've been loaded onto the Pa2x. I was concerened not to raise the noise floor disproportionately by attempting to normalise the pp samples to 100% for example. Do these relative levels for the three sample sets seem reasonable to you?


It's not ideal thing to do but it will work if you have volume on the OSC's set differently. You will likely end up using a pp layer at 127, mf at 90 and ff at 70

Regards.
Sharp.
Back to top
View user's profile Send private message Send e-mail Visit poster's website
mrkorg
Platinum Member


Joined: 03 Nov 2007
Posts: 548
Location: Australia

PostPosted: Sat Feb 14, 2009 3:53 am    Post subject: Reply with quote

Sharp,

What is the advantage of using more than 1 layer?

Does anyone know how many layers used in Grand Piano RX?

Lee,

3 samples per note is incredible. How do you assign 3 samples to one note?
_________________
KORG PA2XPRO
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    Korg Forums Forum Index -> Korg Pa2XPro - Pa800 - Pa500 All times are GMT
Goto page 1, 2, 3, 4, 5, 6, 7, 8  Next
Page 1 of 8

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group